Midterm study guide!
Midterm study guide! Communication Studies 150
Popular in Methodologies in Communication Research
Communication Studies 150
verified elite notetaker
Popular in Communication Studies
This 16 page Study Guide was uploaded by Alyssa Notetaker on Wednesday October 21, 2015. The Study Guide belongs to Communication Studies 150 at University of California - Los Angeles taught by PJ Lamberson in Fall 2015. Since its upload, it has received 125 views. For similar materials see Methodologies in Communication Research in Communication Studies at University of California - Los Angeles.
Reviews for Midterm study guide!
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 10/21/15
Comm Studies 150 Midterm Study Guide Research Design can be messy not a linear stepbystep process like it is below Step one choose a topic 0 Look at your personal interests 0 Look at social problems an issue where something needs to be done better or where there s a societal push to study something 0 Look at the state of the discipline I What s going on in the field of study currently I What people have already done I What people are interested in Have a balance of unique and what s already out there Allows people a frame of reference for your study Most successfulimpactful studies high median conventionality and high tail novelty o Working off of lots of established studies but still have novelty like combining ideas people don t usually combine for example biology and astrophysics combo in communication or something like that 0 Using conventional studies in a new way I Citing mostly conventional studies but also include and connect a few weird outthere studies I Papers with an injection of novelty into an otherwise exceptionally familiar mass of prior work are unusually likely to have high impactquot 0 Look at feasibility I Is it too expensive would take hundreds of years etc Step Two Question Theory and Hypothesis 0 Question 0 Unit of Analysis I The objects that you are studying To find it look at the numbers you re gathering what thing has that number that is associated with the independent variable Ex how does square footage of Walmart affect weekly revenue 0 Unit of analysis store it s the thing that has square footage I Especially focuses on the LEVEL OF RESOLUTION In other words how focused is it Focusing on details or a bigger picture 0 Ex looking at an individual family or company I Why do we care about the unit of analysis We want to avoid incorrect conclusions too broad generalizations fallacy in correlation 0 Variables I what we look at how one variable impacts another variable 0 Theories and Hypotheses I Theory more general I Good theories are Causal o How you think the world works and why 0 We can use correlation to infer causality infer social in uence 0 To do so have to ask two questions I 1 How do we know there are nonrandom clusters In other words is it statistically significant I 2 How do we know the clusters were caused by in uence Have to eliminate o Endogeneity o Confounding variables spuriousness Internally consistent Falsifiable Parsimonious 0 Need control As few moving variables as possible 0 Succinct Predictive o Able to predict outcome I Ex know political orientation so can predict economic status or something like that Terms Unit of Analysis 0 The thing you re studying o What re you looking at What object are you studying an object that stays constant 0 It s the focus of the project Variable o What changes I Ex work habits in summer vs work habits in fall I Unit of analysis is work habits variable season Measurement 0 The process of assigning a numeral or label to the unit of analysis in order to measure it I Ex unit of analysis 5 kids studied over 3 weeks 0 Allows for conceptualization I clarifying what you re going to do I makes it operational makes it something you can actually look at and measure I gives it a complete plan makes it doable Spurious correlation random chance that causes 2 things that are not causally related to be correlated 0 Correlation is just a statistical measurement tells us the numbers we observe tend to move in the same way I Positive correlation when x is high y is likely to be high I Negative correlation when x is low y is likely to be high I Does NOT mean it s causally related In short 3 causal possibilities in uence endogenous or confounding In uence friends in uence friends behavior Endogenous Behaviors in uence connections Confounding neither of the above The Bad News No way to know if the correlation is truly causal in practice because we can never recognize and control for all variables Best way to solve this issue randomized controlled trials Big Data Polls people lie poll could show bias in polling more people from a certain demographic sample bias Data sets can be biased Questions asked can be biased Predicting with web search data 0 Using search data to predict makes the prediction more accurate but only slightly Measurement You have to define it clearly 0 Define subjective and qualitative terms in ways that make it measurable quantifiable Operationalizing Versus conceptualization o Conceptualization is defining and specifying abstract and fuzzy concepts I Concepts have indicators and dimensions Dimensions specifiable aspects of a concept Indicators something the researcher chooses to recognize as a re ection of the variable being studied 0 Makes it more feasible to study something that is not directly observable 0 Ex using smiles to indicate happiness o Operationalization is the development of specific research procedures that will result in empirical observations representing those concepts in the real world I An operational definition specifies precisely how a concept will be measured I Define variables and attributes make them measurable Levels of measurement 0 Binary I Aka dichotomous 0 Nominal I Aka categorical discrete multinomial I Give categories but there s no order to the categories I Ex What religion are youquot Possible answers are 1 Christian 2 catholic 3 protestant 4 hindu 5 muslim 6 Jewish 7 other 8 none I Ex hair color I Ex nationality o Ordinal I Aka categorical discrete multinomial I Give categories but they have an order I Ex satisfaction levels on a scale of 1 to 5 with 1 being dissatisfied and 5 being completely satisfied 0 Interval I Aka continuous can take on any value in a range I Ex measuring cortisol levels temperature placing politicians on a scale of liberal to conservative IQ score You can compare the variables cortisol levels politicians degrees but can t say one is twice the level of the other can t say Hillary Clinton is twice as liberal as Donald Trump You can see the difference between 2 measurements but can t make it a ratio o m I Aka continuous I But in addition to being like interval there is an absolute zero So you can say something is twice as much as something else I Ex wealth measurements age length of time Measurement Quality 0 Reliability I Consistency little variation I Idea that the same data would have been collected each time in repeated observations of the same type I We can increase reliability with increased range of measurement using multiple indicators I Ex a scale is reliable objective opinions on a topic are not 0 Validity I How close is your result to the truth Closer more valid I Valid when the measure accurately re ects the concept it is intended to measure Face validity it makes sense at face value it s reasonable adheres to common sense Criterionrelated validity the degree to which a measure relates to some external criterion Construct validity asks whether the various measures for a given concept all seem to correspond to the same thing Content validity the degree to which a measure covers the concept it operationalizes Error 0 Systematic error I It s systematic the level of the error is always the same I Ex when a scale is off and always adds an ounce to what is being measured I Ex when a ballot is made so it looks like to vote for someone you punch the second hole but really you should punch the third I Sources of systematic error Social desirability 0 Way to minimize this bias I Instead of asking the subject about their own views or supposed actions give a hypothetical situation about a stranger I Ex in this certain situation what do you think this person a stranger madeup person would do Anchoring o The tendency to rely too heavily or anchor on one trait or piece of information when making decisions 0 Usually the subject focuses most on the first piece of information we get from them Acquiescence agree vs disagree 0 People are more likely to agree Order effects 0 If you ask a series of questions or have a list to choose from people are affected by which questions comes first I Can bias the results 0 Random error I Usually cancel each other out 0 Error observed value true value systematic error random error I adjusting the results for error Steps in research review 1 unit of analysis variables and 2 Theory 3 Hypothesis 4 Measurement and operationalization I How to scale what you re measuring I How to quantify it o 5 Operationalizing defining your concept I Is it binary nominal ordinal interval or ratio 0000 Sampling Fundamental property of sampling you infer something bigger from a smaller group test 0 Key characteristics I We want to know about a class of similar objects or events the population I We observe some of these the sample I We form inferences about the population based on the sample 0 Possible errors or bias nonrandom and harder to detect in sampling I Selection bias Undercoverage 0 Some members of the population are inadequately represented in the sample Nonresponse bias 0 Even if it s random sampling people can refuse to answer potentially leading to undercoverage Voluntary response bias 0 People with extreme opinions are more likely to respond I Response bias Leading questions Social desirability 0 Population I Target population the population to which you would like your results to generalize I Sample frame the set of all cases from which your sample is drawn Ideally it matches the target population The characteristics of the target population are called parameters and are usually unknown Sample estimates of population parameters are statistics I Key question to find out if your sample could have selection bias does your sample frame differ systematically from your target population 0 Why sample I We want to be able to draw general inferences about large classes of objects or events but Can t survey a whole target population not feasible Time and money are limited I Sometimes you can get a more accurate result by sampling than by observing the whole population Ex census tries to count EVERYBODY but some people mostly white and upper class were counted twice while even more were not counted at all mostly minorities children and lower income 0 How to sample I 1 Define your population after determining question theory hypothesis units of analysis and measurement target population I 2 Find your sample frame 2 ways to construct a sample frame 0 1 List all cases in the population 0 2 Provide a rule defining membership Find a way to sample the population that does not leave any demographics out disproportionately 0 Ask if your sample frame differs systematically from your target population to avoid selection bias Must have adequate numbers I 3 Choosing your sample You want it to be representative of the population ie proportionally match the target population on key characteristics 0 Problem this is almost impossible to assess we can only measure representativeness with respect to specific characteristics 0 Instead we judge the quality of a sample on how it was obtained I Ways to obtain a sample that avoids bias see types of sampling below 0 Types of sampling Probability sampling All cases in the population have a known probability of being sampled Includes simple random sampling stratified random sampling systematic random sampling cluster random sampling and multistage sampling I Simple random sampling each possible combination of cases has an equal probability of being sampled I Stratified random sample Population is subdivided into exhaustive mutually exclusive categories the strata and you take a simple random sample from each strata Uses I When you want each strata to be homogenous so your sample is heterogeneous includes people from each strata o Increases precision for the sample size I Aka it makes it more efficient don t need to look at so many people I BUT you need to know more about the population like their age and gender which can make it more costly o Allows the researcher to highlight a specific subgroup within the population it ensures the presence of the key subgroup in the sample I Cluster Sampling Breaks the population into clusters Select at random a sample of clusters Obtain a full list of the cases within the sampled clusters Sample from within the clusters This is good for when things are basically the same across the clusters I Systematic sampling Sample every kth case the number k the sampling interval 0 Ex choose every 5th person on a list Issue with this bias 0 Ex sampling every 1St and 5th house on the block I Can be all corner houses with bigger lots and bigger houses 0 Ex every 10th person on a military list I But every 10th person was a sergeant o Nonprobability sampling I Does not control for investigator bias I Statistical properties unknown cannot use theory of random sampling to statistically eliminate bias I Types Convenience sampling Purposive sampling Quotas o Stratified random sampling where there are quotas for number of people in each strata Referral or snowball sampling 0 Good for small populations 0 Survey someone then they refer you to another 0 Good for stuff like finding gang members to question no official list of the population 0 Sample error I How far away the number you estimate in the sample is from the truth I As the sample size increases sample error decreases It gets more accurate with a larger sample The distribution is less varied and more centered on the true value 0 So you want a larger sample size especially when the population is extremely varied heterogeneous o If your sample distribution is really varied it s a cue that you need a bigger sample size I Standard error The average sample error of the sampling distribution The standard deviation of the population divided by the square root of the sample size I 9 Confidence intervalquot or the range of values within the estimated population value is likely to lie We are 95 confident the population mean is 3 i 1 Experimentation Basic characteristics 0 True experiments eliminate effect of exogenous variables and allow strong causal inferences eliminate confounders and reverse causality endogeneityD o Tests causal relations I A manipulated independent variable is followed by a measured dependent variable Testing the effect of one variable on another I There are two groups a control and a treatment group 2 groups treated exactly the same except for the one manipulation I In this way we can test for causality and avoid making conclusions about something that actually just happened randomly Inferring causality all the below have to happen be checked 0 1 Association correlation o 2 Direction of in uence ie not endogenous o 3 Elimination of rival explanations ie not spurious or confounded I In an airtight experiment only one rival explanation chance Matching v random assignment 0 Matching matching subjects on characteristics that logically seem to be related to the experimental outcome and putting one subject in the test and one in the control group 0 Random completely random or SRS stratified etc Measurement validity o The way you operationalize the experiment must capture what you want to study Evaluating internal validity 0 Random assignment 0 Manipulation of independent variable 0 Measurement of the dependent variable 0 At least one control group 0 The constancy of conditions across groups except control External validity o Is it generalizable Staging an experiment 0 Pretesting I Train the quotcastquot the interviewers testers I Test the instructions cover story etc I Check if the manipulation has the desired effect I Revise and practice the quotscriptquot I See if the subject will remain detached or if they ll experience experimental realism experience it like it s real life and so re ect the truth Another level of realism mundane realism similarity of experimental events to everyday experiences Demand characteristics cues telling subjects what is expected of them and what the experimenter hopes to find Affects that happen because the subject knows they re being watched reactive measurement effects 0 Sometimes there s evaluation apprehension where the subject acts in a way to try to get a favorable evaluation of the experimenter 0 Sometimes want to quothelpquot the experimenter or give intentionally useless or invalid responses I Something else to keep an eye out for experimenter effects Subtle effects due to how the experimenter believes the experiment will turn out They may unintentionally communicate to subjects how they should respond to confirm a hypothesis Includes testing bias a pretest can affect what the subject notices in the experiment prime them for the actual experiment I Minimizing these experimental biases Ask the subject about their perception of the experimental environment whether it seemed like they should act a certain way etc Doubleblind technique 0 Subject recruitment I Acquisition of informed consent 0 Introduction to experiment I Sometimes includes a cover story about what the study is for to avoid bias while allaying preoccupation with what it s about Random assignment Manipulation of independent variable Measuring of dependent variable Manipulation check I See if any result actually happened if the manipulation caused any affect on the subject 0 Debriefing I Discuss what happened with the subject Experimentation outside the laboratory 0 Field experiments I a true experiment in a natural setting 0 Experimental design in survey research 0 Using units of analysis other than individuals OOOO Threats to internal validity Testing bias 0 The way in which the measurement is taken changes the response 0 When the test experiment itself in uences the subjects responses 0 Ex a pretest primes subjects as to what the researcher is looking for in uences how they act in the experiment 0 Similar to experimental bias demand characteristics reactive measurement effects Instrumentation o The way in which the measurement is taken changes over time I Unwanted changes in characteristics of the measuring instrument or in the measurement process 0 Can get more accurate less accurate change qualitative standard standard of good vs bad changes as you see better or worse examples I Ex A person hired to count events gets bored and less accurate as time passes I Ex someone gets better at measuring as time goes on I Ex a professor grading exams Selection bias when there are systematic differences between the composition of the control and treatment groups 0 Ex selfselection selfreporting people report if they listen to radio or watch TV then are asked to answer some questions the results could be because more politically inclined people choose to listen to radio or something and not because listening to the radio causes them to be more politically inclined endogeneity o Confounders lack of causation I Ex more accidents where there are crosswalks confounder more people use crosswalks than not and more dangerous streets have crosswalks so should not conclude that crosswalks cause accidents 0 Ex people who sign up for the treatment it s not random 0 Avoid with random assignment History if something is widely known people will be affected by it 0 When other events in the environment might affect the outcome 0 Ex if you do an experiment over 2 days and the people who do it the first day talk to the people who will do it the second day Maturation 0 When changes take place within subjects over time during the course of the experiment I Ex if the experiment goes on for a long time the subject can get tired bored hungry etc Statistical Regression 0 Extreme measurements tend to move closer to the mean on second observation 0 Ex people who score low on a pretest are likely to score closer to the average on a posttest I Taking a test primes the subject for taking tests so they do better I Higher scores are likely to decrease on the second test I So never use a test to assign people to treatments I Keep assignment random Attrition the loss of subjects in an experiment 0 Greatest threat when there is differential attrition when the conditions of the experiment the test group 1 2 control etc have different dropout rates 0 Can confound the experiment Preexperimental Designs Use debaters from one party to reduce variation Use matching to randomly assign the same number of each democrats and republicans to each TV and radio 1 Oneshot case study some treatment is administered to one group after which the group is observed or tested for treatment effects 0 No control or second test group possibility for attrition maturation history and other confounders 2 Onegroup pretestposttest design 0 Observe a group of subjects pretest introduce the treatment experiment with independent variable observe posttest o Allows comparison controls for attrition checks who was there before and after 0 No control possibility for testing bias as the pretest could affect the subjects views of what s happening also maturation history instrumentation statistical regression 3 The static group comparison 0 Two groups one receives treatment and the other doesn t both are observed afterward posttest 0 Has a control Also controls for history because both groups should have pretty much the same history and for testing and statistical regression since it doesn t have a pretest 0 Still threats to internal validity selection not necessarily randomized attrition maturation if not done simultaneously True Experimental Designs 4 The pretestposttest control group design 0 Randomly assign subjects to 2 groups observe both pretest apply treatment to one observe both posttest 0 External validity threat testing interacting with the independent variable aka testingtreatment interaction in other words the results of the treatment could be different than they would be without a pretest 5 The posttestonly control group design 0 Randomly assign the subjects to 2 groups treat one observe results of both 0 This is more economical and eliminates the possibility of testingtreatment interaction 0 But chance of attrition if it s a long term experiment 6 The Solomon fourgroup design 0 Randomly assign subjects to 4 groups I Observe pretest only two of them I Test apply the treatment to one of the pretested groups and one of the nonpretested groups I Observe all groups posttest compare to see the effect of the treatment and the effect if any the pretest had Withinsubjects design have both treatments apply to the same subject 0 Don t have to use so many people reduce issues of individual differences some are more informed from different parties etc o Comparing someone to themself I each person acts as their own control 0 Issues I Subject can be affected by which goes first Solution test if there is an order effect by randomly assigning half to see TV first half to hear radio first Counterbalancing I The subject can figure out what the researcher is studying Hiding the real experiment to simulate reality avoid experiment effects experimental bias where people change their behavior because they know they re being tested Factorial Experimental Designs Social phenomena are often caused or in uenced by multiple variables So factorial designs allow researchers to study several independent variables at once Some types of factorial designs 0 2x2 factorial design is when there are 2 factors independent variables each with 2 levels I Ex the Solomon fourgroup design 2 independent variables pretest and treatment 2 levels each whether or not the group is pretested whether or not the group is treated 4 total conditions pretest treated pretest not treated no pretest treated no pretest not treated I Has 4 total conditions 0 3x4 factorial design 3 independent variables each with 4 levels I 12 conditions Factorial designs provide information about the main effect of each factor 0 Aka the overall effect of the factor by itself 0 Found by comparing the overall means of each condition I Ex compare the overall means of the treated groups and the non treated groups to see the effect of the treatment I Compare the overall means of the pretested group and the non pretested group to see the effect of the pretest Interaction Effects Gives us information about the joint effects of the factors together for example the treated and pretested vs the nontreated and nonpretested To see visually graph the means of each condition by factor 14 12 Task 1 Task 2 Control Treatment 0 Condition 0 Each line represents a condition for example blue shows pretested and red shows nonpretested and the nontreated and treated points are on the x axis I How it is shown the treatment affects the subjects in all cases but more so when the subjects were pretested I If the red line were at it would show the treatment had no effect and all the seen effects were due to the pretest I If the blue line were at it would show the pretest had no effect I If they re parallel it shows they had no effect on each other no interaction Quasiexperimental designs Separatesample pretestposttest design 0 Randomly assign subjects to two groups observe one pretest test both observe the other one posttest 0 Most serious threat to validity history Nonequivalent control group designs 4 O Performance D N A m 03 0 Where the researcher uses already intact groups like school classes 0 So random assignment is impossible Interrupted timeseries design 0 Multiple observations before and after the treatment 0 Threat from history instrumentation if you re looking at records and there has been a change in recordkeeping procedures 0 This with 2 groups one with and one without treatment multiple timeseries design Surveys 3 Stages to surveys after hypothesis operationalization etc Sample Survey Analyze Conversational norms Our norms can affect how someone takes or understands a survey 0 Be truthful 0 Be relevant on topic 0 Don t repeat 0 Be clear So if we ask two synonymous questions that are worded differently Considerations Open v closed questions 0 Open more information more open for interpretation allows the person to submit whatever answer is most relevant to them I Can allow for more information doesn t enforce your own biases on what s important You can find out if you re looking in the right direction I To get a quantifiable answer coding Have predetermined categories of answers and after the survey categorize each person s answer into one of the categories Possible to automate the coding process set a computer to find certain phrases and automate the counting of the types of each response 0 Closed have to answer with a menu of options can give a very precise answer guarantees a certain type of answer Easier to quantify Direct v indirect 0 Direct ask exactly what you want to know 0 Indirect ex do a hypothetical situation between people who are not the interviewee I if someone is invited over to Net ix and chill with someone else how likely do you think it is they will sleep together I Best to use when there s a chance of bias criminal activity touchy subjects etc Response format yesno categorical numeric scale etc Visual aids 0 Be careful of inadvertently introducing a picture that would in uence people to give a certain answer 0 Ex if we want to know how healthy someone is as in not sickly we wouldn t want to use a picture of someone working out which would have people associate the quiz with how athletically healthy they are 0 The visuals can also clarify what type of answer you re looking for Question order Rules of thumb for questions Keep it simple and clear Address one issue per question Don t ask loaded questions ex Don t you agree thatquot Avoid emotionally charged language Avoid double negatives
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'